The present invention relates to a low latency lateral decoding of data stored on magnetic tape, and more particularly, to decoding of data stored using lateral encoding.
Error correction coding (ECC) in linear tape open (LTO) and enterprise tape drives is used to ensure that the byte error rate (BER) at the output of the ECC decoder is less than about 1×10−19 even though the BER at the output of the detector may be around 1×10−3. Furthermore, ECC is able to correct large spatial errors on magnetic tape. In particular, ECC is able to correct some large burst errors due to media defects. It is also capable of correcting all errors in a lateral tape stripe of large width which may occur due to instantaneous speed variations during tape transport from one reel to the other reel of a tape cartridge. Finally, ECC is able to correct a large number of dead tracks due to, e.g., temporarily or permanently non-functioning reader elements in the magnetic tape head of the tape drive. Modern tape drives are able to correct M/8 dead tracks, where M is the total number of tracks that are simultaneously written onto and read from the magnetic tape.
In tape storage, relatively long (about 1 kB) longitudinal interleaved error correction codewords, which are commonly referred to as codeword interleaves (CWI-4) consisting of four byte-interleaved row codewords from four different product codewords, are written on tracks of the magnetic tape. In current receiver architectures that perform a single pass through the digital front-end functions and detection, the decoding latency associated with read codewords is prohibitively large for decoded bits to be used to drive decision-directed digital front-end functions. Clearly, large delays in decision-directed loops are not desirable and therefore should be avoided. Such a scheme would introduce an unmanageable amount of latency to the reading and decoding process in current tape drive read architectures. In receiver architectures that iterate between decoder and digital front-end functions, the buffer size used to store the samples at the output of the analog-to-digital (A/D) converter is very large when decoded bits in current tape ECC architectures are used to drive decision-directed digital front-end functions during a second and subsequent iterations. Furthermore, the error correction capability of these longitudinal codewords in current tape drives is relatively weak.